As a double major attending a gravitational waves focused REU at Penn State was the necessary push I needed to focus on physics more than computer science.
I got an opportunity to present my work for the first time in front of a board and they extended my research for a of couple months because they were happy. Although just starting this has been the coolest thing to happen to me career wise
Applied for my first student research position. I've only ever been a cook. It's scary to take on less income that ends in four months but I know this will help me land a more permanent and well paying position. Finding out my mom had cancer just made me want to make her proud.
@iharabukhouski A self improving design is the path forward. But the self improving models I see now are having another agent come in and say hey you're doing this wrong. So then we're in this loop where you are attempting to improve based on suggestions of the thing you're trying to improve on
@iharabukhouski However for the AI system to go beyond our knowledge it would need to perceive universal raw data and make it's own interpretation. Maybe it could surpass human intelligence but I believe it would limited as far as being the smartest "human".
@iharabukhouski I want to say no. I used to say yes. Any tech built off our understanding of math and physics is limited by our window of perception. I theorize AI systems will only be highly trained on our knowledge and maybe it could generate something new based on patterns we didn't realize.
Off to Tokyo for #ASSC27. Never been to this meeting before but looking forward to giving a talk about the cognitive implications of a developing self in human infants. https://t.co/rLFwna9CNp
@pranavishnoi @ylecun I'm reminded of Andrej Karpathy's breakdown of LLM's in which I'm paraphrasing but he says the compression is lossy and not lossless. To me it's a world model based on it's goals which is next best token, not adhering to physical laws if necessary.
@ylecun I'm an undergraduate student double majoring in computational physics and undecided on C.S., C.S./E, or C.E.
Do you think AGI will be achievable through software innovation alone? Or rather will we need a new compute structure such as quantum or ternary?
Came in first at my school science fair and second in the computer science fair. I started analyzing cat meows. I got a basic CNN to classify if a cat meow is in relation to food with 76% accuracy. Limitations such as a small dataset will be addressed in future iterations.
Just got @openchatdev 's `openchat/openchat-3.5-1210` running in LibreChat UI!
thx to @AlpayAriyak for the backend assistance, & @runpod for the compute. Also thx to @lgtm_hbu et. al. for Librechat.
Excited to work on this some more over the next few weeks.